When AI designs a drug, who gets the credit?
Insilico Medicine's generative AI platform proposed a drug for pulmonary fibrosis, yet the patent lists only human inventors, highlighting the legal barrier that AI cannot be named an inventor US courts have ruled that "inventor" under patent law means a human being, as established in the DABUS test case brought by attorney Ryan Abbott The US Patent and Trademark Office has shifted from requiring disclosure of AI involvement to treating AI as merely a tool, like a calculator Legal experts warn t
Analysis
TL;DR
- Insilico Medicine's generative AI platform proposed a drug for pulmonary fibrosis, yet the patent lists only human inventors, highlighting the legal barrier that AI cannot be named an inventor
- US courts have ruled that "inventor" under patent law means a human being, as established in the DABUS test case brought by attorney Ryan Abbott
- The US Patent and Trademark Office has shifted from requiring disclosure of AI involvement to treating AI as merely a tool, like a calculator
- Legal experts warn that if AI-generated outputs cannot be patented, it could stifle innovation in AI-driven drug development
- The core legal question is evolving: not whether AI can invent, but whether any human contributed enough to qualify as a co-inventor
Why It Matters
This article exposes a critical legal bottleneck at the intersection of AI and intellectual property that directly affects biotech and pharma companies investing heavily in AI-driven drug discovery. As AI systems become capable of generating novel chemical structures with minimal human intervention, the current patent framework—which requires human inventorship—creates uncertainty around IP protection, potentially undermining the economic incentives for AI innovation in drug development.
Technical Details
- Insilico Medicine used its generative AI platform to propose a drug molecule for pulmonary fibrosis, claiming AI "discovery" in press materials but listing five humans (including CEO Alex Zhavoronkov) as inventors on the patent
- The DABUS test case, brought pro bono by Ryan Abbott, challenged whether an AI could be named inventor of a heat-transfer food container design; the DC Appeals Court ruled in 2022 that "individual" in US statute means a human being
- The USPTO's stance has oscillated: under the Biden administration, guidance was published to help applicants assess human co-inventorship in AI discoveries; under Trump, the policy reversed to treat AI as a tool with no disclosure required
- Sarah Korman of Isomorphic Labs (Alphabet spinout) emphasized that "there needs to be a human inventor or there's no invention and no patent," while acknowledging laws must evolve
- Abbott warns that incorrect inventorship listing could become a pathway to invalidate AI-generated drug patents, and questions whether "pushing a button" on an AI should qualify as inventorship
Industry Insight
- AI drug discovery companies should maintain rigorous documentation of human involvement in the invention process to safeguard patent eligibility, as the threshold for qualifying as a co-inventor remains legally ambiguous
- The current "don't ask, don't tell" USPTO approach creates short-term convenience but long-term IP risk; companies should prepare for potential legal challenges to inventorship as AI systems require less human oversight
- Stakeholders should monitor legislative and judicial developments closely, as the tension between AI's growing inventive capacity and human-only patent law will likely produce landmark cases that redefine IP frameworks for the AI era.
Disclaimer: The above content is generated by AI and is for reference only.